Summary of delisting model results

In this chapter the best performing delisting model performances are summarized for each testing set.

Table 28: The evaluation metrics of best performing delisting predictive models for each testing set.

Model data set TP FP TN FN Precision Recall Accuracy MCC SVM_L_DE3 DE1_test 6 35 115 5 0.146 0.545 0.752 0.181 SVM_L_DE3 DE2_test 10 31 112 8 0.244 0.556 0.758 0.245 SVM_L_DE3 DE3_test 10 31 107 13 0.244 0.435 0.727 0.169

Table 28 summarizes the best performing models on each market switching testing set. By examining the results of the best performing models we can conclude that the linear kernel MS3 SVM model was the best performing predictive model on all testing sets. Noteworthy from the results is that even the model predicted the most false positives the model did not predict more true positives between testing set 2 and testing set 3 even though the testing set 3 has 5 more delisting event observations. From the results we can also notice that the models produced many misclassifications by predicting over 30 false positives in each training set. Also in each training set, there were few false negatives predicted. Based on MCC metric the best performance with the model was achieved on testing set 2 in which the events are occurring within two years. From the results we can conclude that even though the model could correctly predict portion of the event observations, there were still misclassifications. This can be caused by the differing characteristics between involuntary and voluntary delisting, which were not separated in the training or testing sets. From the economic usability perspective the models could be used as an early warning guidance.

8 DISCUSSION

The explanatory analysis and the variable / feature plane analysis supported the existing theories and empirical studies conducted on both cases; market switching and delisting events. In the market switching model building and analysis the most interesting variable included to the model came from the indicator category where the years listed on First North -variable. The variable feature plane indicated that the companies are more likely to switch markets around year three. This effect has not been indicated in any other studies reviewed. It should be noted that the market switching events have attracted only low amounts of academic interest and most relevant studies have focused more on cross-listing events. The years listed on First North -variable also indicated that delisting events are more likely to occur after the company has been listed for three years. The delisting variables were from the other parts in line with earlier studies conducted especially from the involuntary delisting perspective. Overall market switching events have attracted only a little academic research and thus some of the variables included in the explanatory analysis and in the models could be seen contributing in the examination of the characteristics of market switching companies.

In the prediction of market switching events the results suggest that by using the variables based on the theoretical framework and explanatory analysis it is possible to identify market switching events. The best results were obtained for the two year prediction horizon, but it should be noted that even though some of the events were correctly predicted, there were also misclassifications by classifying a non-event as an event and vice versa. The methodology used in the study contributes to the existing studies also, as the methodologies have not been used in the study of market switching events based on the literature review on the existing studies. The prediction can also be biased as, there were generated values and observations within the training sets used but based on the analysis of the datasets we can conclude the effects to be marginal.

The delisting results were less mixed from the model performance perspective.

The best performing model on all datasets was the support vector machine model with linear kernel. From the results we can conclude that even though the models have some predictive power it still misclassified some of the delisting events and thus based on the model performance they should be considered to be used as guiding principles and not as exact predictions. From the perspective of model performance, there were several possible factors influencing the models performance. First, in the study involuntary delisting and voluntary delisting events were treated as the same and thus it may have worsen the performance due to differences in characteristics of the companies in different types of delisting events. Secondly, there were generated values and observations in the training sets, which can also cause bias in the performance of the models.

9 CONCLUSIONS

This study examines the possibility of the use of public financial statement- and market information in the prediction of market switching and delisting events from OMX First North Nordic multilateral exchange. The data used in this study consists of companies listed in OMX First North Nordic in timeframe beginning from 2007 extending to 2012. Each firm year observation was treated as an individual observation. The study was conducted in three stage approach on which first relevant theoretical framework were constructed for market switching and delisting events. Based on the constructed theoretical framework initial predictor pool was gathered. In the second stage, explanatory analysis was conducted for the initial variable pool in order to further study the characteristics of market switching and delisting companies. In addition, the examination of characteristics the variable scope for predictive models was narrowed on the basis of theoretical base and explanatory analysis. Based on the narrowed variable group predictive models were constructed and trained using three different training sets in order to examine the possible noise in the dataset. In the third stage, the trained models were used to predict on three different testing sets on which the event horizon varied from within one year to three years.

Based on the empirical evidence of the study we can state, that explanatory analysis was able to identify characteristics influencing on market switching events. The findings contribute new variables for academic research and found new support for earlier characteristics identified in empirical and theoretical research of market switching events. The empirical results also support the usability of financial statement and market information in the prediction of market switching events. The empirical results indicate that prediction of the market switching events is possible by using the chosen methodology and data to some extent. The results acquired in this study may function as a baseline for OMX First North Nordic market switching event research, as only few

studies have been conducted for the market in the narrow research field of market switching event prediction.

The empirical evidence support the viability of the use of self-organizing maps methodology in the identification of theoretically relevant variables. This also provides evidence for theoretical research and support the empirical findings of earlier research. In the explanatory analysis also new significant variables were identified. The prediction of delisting events using the three stage approach used in the study proved to be able identify portions of the delisting events in the testing sets. Thus it can be concluded that financial statement and market information can be used in the prediction of delisting events by the chosen methodology. It has to be noted that even though the models were able to identify portion of the delisting events also misclassifications were predicted by the models. The empirical findings of this study can be used as baseline performance as the market and delisting events as a topic have attained only little academic interest thus far.

One of the major limitations of the study is caused by the nature of the studied data as the main focus of the study is on companies, which switched market or delisted from OMX First North Nordic. First limitation is caused by the disclosure requirements posed by the multilateral exchange, which are not as strict as in main exchanges, causing the data gathering process from accessible databases to be challenging. Secondly majority of the data set consist of delisted companies some of which have delisted involuntarily from the exchange, due to bankruptcy or merger/acquisition the financial statement data was not available for all companies in the chosen sample. Thirdly, there were challenges in the data collection process as some of the companies changed name during the observation period. One limitation posed for the study was the rare nature of the event and missing values due the nature of the events. This caused the necessity to use imputation methodology for missing values and data generation for training sets in order to compensate for

the missing values and small data sets with highly imbalanced classes. Also the performance of the delisting predictive models could have been affected by not differentiating between involuntary and voluntary delisting events.

It should also be noted that the chosen time period includes the subprime crisis and the Eurozone crisis. The crises can affect the amount of events occurring during the period. It can be assumed that the crises have increased slightly the amount of delisting events, due to the direct and indirect effect on financial stability of companies, and it could also have skewed the ratio between involuntary and voluntary delisting events towards involuntary delisting. Market switching events are affected in inverse manner. This could have an effect on the models generalization in time periods not affected by the crises by raising the amount of delisting events predicted by the models and decreasing the amount of predicted market switching events.

For further study the base variable set could be expanded by including more growth rates, which could be used to capture the movement of the observations in the variables chosen. Also sector indicators could provide information about differences between sectorial behaviour relating to market switching or delisting events. Furthermore different methodologies could be used in order to identify the best methodology for the prediction of the events. Moreover it could provide more insight for the nature of the events if the models could be tested on several multilateral stock exchange data. This would assist the generalization of the acquired results.

REFERENCES

Amihud, Y. & Mendelson, H., 1986. Asset pricing and the bid-ask spread.

Journal of Financial Economics, Volume 17, pp. 223-249.

Angel, J., Harris, J., Panchapagesan, V. & Werner, I., 2004. From Pink Slips to Pink Sheets: Liquidity and Shareholder Wealth Consequences of Nasdaq Delistings. Working Paper Series from Ohio State University, Charles A. Dice Center for Research in Financial Economics, pp. 1-49.

Aziz, M. & Dar, H., 2004. Predicting Corporate Bankruptcy: whither do we stand?. Working paper.

Bailey, W., Karolyi, G. & Salva, C., 2006. The economic consequences of increased disclosure: Evidence from international cross-listings. Journal of financial economics, 81(1), pp. 175-213.

Baiman, S. & Verrecchia, R., 1996. The Relation among Capital Markets, Financial Disclosure, Production Efficiency, and Insider Trading. Journal of Accounting Research, 34(1), pp. 1-22.

Baker, K. & Edelman, R., 1990. OTC Market switching and stock returns: Some empirical evidence. The Journal of Financial Research, 13(4), pp. 325-338.

Bancel, F. & Mittoo, U., 2001. European managerial perceptions of the net benefits of foreign stock listings. European financial management, 7(2), pp.

213-236.

Bellovary, J., Giacomino, D. & Akers, M., 2007. A Review of Bankruptcy Prediction Studies: 1930 to Present. Journal of Financial Education, Osa/vuosikerta 33, pp. 1-42.

Benos, E. & Weisbach, M., 2004. Private benefits and cross-listings in the United States. Working paper, pp. 1-30.

Breiman, L., 2001. Random Forests. Machine learning, 45(1), pp. 5-32.

Breiman, L., 2003. Random Forests. [Online]

Available at:

http://www.stat.berkeley.edu/~breiman/RandomForests/cc_home.htm [Haettu 12 2 2014].

Bris, A., Cantale, S. & Nishiotis, G., 2007. A breakdown of the valuation effects of international cross-listing. European financial management, 13(3), pp. 498-530.

Caruana, R. & Niculescu-Mizil, A., 2006. An empirical comparison of supervised learning algorithms. New York, ICML '06 Proceedings of the 23rd international conference on Machine learning, pp. 161-168.

Cetorelli, N. & Peristiani, 2010. Firm value and cross-listings: the impact of stock market prestige. FRB of New York staff Report, Issue 474, pp. 1-45.

Chaplinsky, S. & Ramchand, L., 2007. From Listing to Delisting: Foreign Firms’

Entry and Exit from the U.S.. Working paper.

Charitou, A., Louca, C. & Panayides, S., 2007. Cross-listing, bonding hypothesis and corporate governance. Journal of business finance&

accounting, 34(7), pp. 1281-1306.

Chawla, N., Bowyer, K., Hall, L. & Kegelmeyer, W., 2002. SMOTE: Synthetic Minority Over-sampling Technique. Journal of Artificial Intelligence Research, 16(1), pp. 321-357.

Chemmanur, T. & Fulghieri, P., 2006. Competition and cooperation among exchanges: A theory of cross-listing and endogenous listing standards. Journal of Financial Economics, 82(2), pp. 455-489.

Chen, K. & Schoderbek, M., 1999. The role of accounting information in security exchange delisting. Journal of Accounting and Public Policy, 18(1), pp.

31-57.

Cissé, A. & Fontaine, P., 2013. Why do companies transfer the trading compartment of their common stocks. Working paper, pp. 1-22.

Coffee, J., 2002. Racing towards the top?: The Impact of Cross-Listings and Stock Market Competition on International Corporate Governance. Columbia Law and Economics Working Paper, Volume 205, pp. 1-88.

Cortes, C. & Vapnik, V., 1995. Support-Vector Networks. Machine Learning, 20(3), pp. 273-297.

Cowan, A., Carter, R., Dark, F. & Singh, A., 1992. Explaining the NYSE Listing Choices of NASDAQ Firms. Financial Management, 21(4), pp. 73-86.

DeAngelo, H., DeAngelo, A. & Rice, E., 1984. Going Private: Minority Freezeouts and Stockholder Wealth. Journal of Law and Economics, 27(2), pp.

367-401.

Djama, C., Martinez, I. & Serve, S., 2012. What do we know about delistings?

A survey of the literature. Working paper, pp. 1-22.

Doidge, C., Karolyi, A. & Stulz, R., 2004. Why are foreign firms listed in the U.S.

worth more?. Journal of financial economics, 71(2), pp. 205-238.

Doidge, C., Karolyi, A. & Stulz, R., 2010. Why do foreign firms leave U.S. equity markets?. The Journal Of Finance, 65(4), pp. 1507-1553.

Foester, S. & Karolyi, A., 1999. The effects of market segmentation and investor recognition on asset prices: Evidence from foreign stocks listing in the United States. The journal of finance, pp. 981-1013.

Fürst, O., 1998. A theoretical analysis of the investor protection regulations argument for golbal listing of stocks. Working paper, pp. 1-21.

Goto, S., Watanabe, M. & Xu, Y., 2009. Strategic disclosure and stock returns:

Theory and evidence from US cross-listings. The review of financial studies, 22(4), pp. 1585-1620.

Hail, L. & Leuz, C., 2004. International differences in the cost of equity capital:

Do legal institutions and securities regulation matter?. Workin paper, The Wharton school .

Hail, L. & Leuz, C., 2005. Cost of Capital and Cash Flow Effects of U.S. Cross-listings. Working paper, pp. 1-43.

Halling, M., Pagano, M., Randl, O. & Zechner, J., 2008. Where is the market?

Evidence from Cross-listings in the U.S.. Review of financial studies, 21(2), pp.

725-761.

Jenkinson, T. & Ramadorai, T., 2013. Does one size fit all? The consequences of switching markets with different regulatory standards. European Financial Management, 19(5), pp. 852-886.

Jensen, M., 1986. Agency costs of free cash flow, corporate finance, and takeovers. The American Economic Review, 76(2), pp. 323-329.

Kadlec, G. & McConnell, J., 1994. The effect of market segmentation and illiquity on asset prices: Evidence from exchange listings. The Journal of Finance, 49(2), pp. 611-636.

Karatzoglou, A., Meyer, D. & Hornik, K., 2006. Support Vector Machines in R.

Journal of Statistical Software, 15(9), pp. 1-28.

Karolyi, A., 2003. The role of ADRs in the development of emerging equity markets. Working paper, pp. 1-52.

King, G. & Zeng, L., 2001. Logistic Regression in Rare Events Data. Political Analysis, 9(2), pp. 137-163.

Kohonen, T., 1982. Self-Organized Formation of Topologically Correct Feature Maps. Biological Cybernetics, 43(1), pp. 56-69.

Kohonen, T., 1998. The self-organizing map. Neurocomputing, 21(3), pp. 1-6.

Königsgruber, R., 2009. An economic analysis of cross-listing decisions and their impact on earnings quality. Schmalenbach Business Review, Volume 61, pp. 310-330.

Lang, M., Lins, K. & Miller, D., 2003. ADRs, Analyst, and Accuracy: Does Cross listing in the United States improve a Firm´s Information Environment and increase market value?. Journal of Accounting Research, 41(2), pp. 317-345.

Lasfer, M. & Pour, E., 2012. The Impact of Leverage on the Delisting Decision of AIM Companies. Working paper.

Lehn, K. & Poulsen, A., 1989. Free Cash Flow and Stockholder Gains in Going Private Transactions. The Journal of Finance, 44(3), pp. 771-787.

Liaw, A. & Wiener, M., 2002. Classification and Regression by randomForest.

R News, 2(3), pp. 18-22.

Licht, A., 2003. Cross-listing and corporate governance: Bonding or avoiding?.

Chicago Journal of International Law, Volume 4.

Li, J., Zhang, L. & Zhou, J., 2005. Earnings Management and Delisting Risk:

The Case of IPO Firms. Working paper.

Maimon, O. & Rokach, L., 2010. Data Mining and Knowledge Discovery Handbook. 2nd toim. New York: Springer.

Martinez, I. & Serve, S., 2011. The delisting decision: the case of buyout offer with squeeze-out (BOSO). International Review of Law and Economics, 31(4), pp. 229-239.

Melvin, M. & Valero, M., 2009. The dark side of international cross-listing:

effects on rival firms at home. European financial management, 15(1), pp. 66-91.

Merton, R., 1987. A Simple Model of Capital Market Equilibrium with Incomplete Information. The Journal of Finance, 42(3), pp. 483-510.

Min, J. & Lee, Y., 2005. Bankruptcy prediction using support vector machine with optimal choice of kernel function parameters. Expert Systems with Applications, 28(4), pp. 603-614.

Mitchell, T., 1997. Machine learning. 1 toim. Boston: MIT Press and The McGraw-Hill Companies, Inc..

Mittoo, U., 1992. Managerial perceptions of the net benefits of foreign listing:

Canadian evidence. Journal of international financial management and accounting, 4(1), pp. 40-62.

NASDAQ OMX First North, 2014. First North Nordic – Rulebook. [Online]

Available at: http://www.nasdaqomx.com/digitalAssets/89/89108_first-north-rulebook-2014-01-01.pdf

[Haettu 2 2 2014].

NASDAQ OMX Helsinki Ltd., 2013a. Disclosure rules for First North Finland

Premier - segment. [Online]

Available at: http://www.nasdaqomx.com/digitalAssets/87/87447_disclosure-rules-for-first-north-finland-premier-segment-1-july-2013-final.pdf

[Haettu 2 2 2014].

NASDAQ OMX Helsinki Ltd., 2013a. Rules of the exchange. [Online]

Available at: http://www.nasdaqomx.com/digitalAssets/87/87944_rules-of-the-exchange_02102013.pdf

[Haettu 1 2 2014].

NASDAQ OMX Nordic, 2014a. STATISTICS. [Online]

Available at:

http://www.nasdaqomxnordic.com/uutiset/tilastoihin?languageId=1&Instrumen t=SSE101

[Haettu 4 2 2014].

NASDAQ OMX Nordic, 2014b. MARKET NOTICES FIRST NORTH. [Online]

Available at:

http://www.nasdaqomxnordic.com/uutiset/markkinapaikan_ilmoituksia/firstnort h

[Haettu 10 1 2014].

NASDAQ OMX, 2013. Guide to listing on NASDAQ OMX FIRST NORTH.

[Online]

Available at: http://www.nasdaqomx.com/digitalAssets/85/85721_q13-0970firstnorthlistingguide_final_may21.pdf

[Haettu 4 2 2014].

NASDAQ OMX, 2014a. ABOUT US. [Online]

Available at:

http://www.nasdaqomxnordic.com/about_us?languageId=1&Instrument=SSE 101

[Haettu 4 2 2014].

NASDAQ OMX, 2014b. Differences in Admission Criteria for the Main Market

and NASDAQ OMX First North. [Online]

Available at:

http://www.nasdaqomx.com/listing/europe/primarylisting/differencesinadmissi

oncriteria

[Haettu 5 2 2014].

Ng, Y., Yong, H. & Faff, R., 2012. The long- and short-run financial impacts of cross listing on Australian firms. Australian Journal of Management, 38(1), pp.

81-98.

Pagano, M., Röell, A. & Zechner, J., 2002. The geography of equity listing: why do companies list abroad?. The journal of finance, 47(6), pp. 2651-2694.

Pestian, J. & Matykiewicz, P., 2012. Effect of small sample size on text categorization with support vector machines. Montréal, Association for Computational Linguistics.

Philips, A., 1988. Stock Market Values And P/E Ratios As Predictors Of New York Stock Exchange Delistings. The Journal of Applied Business Research, 4(3), pp. 56-65.

Pour, E. & Lasfer, M., 2013. Why Do Companies De-List Voluntarily from the Stock Market?. Journal of Banking & Finance, 37(12), pp. 4850-4860.

Quinlan, J., 1986. Induction of decision trees. Machine Learning, 1(1), pp. 81-106.

Ravi, P. & Ravi, V., 2007. Bankruptcy prediction in banks and firms via statistical and intelligent techniques - A review. European Journal of Operational Research, 180(1), pp. 1-28.

Reese, W. & Weisbach, M., 2002. Protection of minority shareholder interests, cross-listing in the United States, and subsequent equity offerings. Journal of financial economics, Volume 66, pp. 65-104.

Rokach, L. & Maimon, O., 2005. Top-down induction of decision trees classifiers - a survey. IEEE Transactions on systems, man and cybernetics:

part C, 35(4), pp. 476-487.

Ruiz, I., Riportella, C. & Requejo, S., 2014. SMEs’ Delisting Decisions on the Alternative Investment Market (AIM): Family Holders and Financial Crisis.

Working paper.

Sanger, G. & Peterson, J., 1990. An empirical analysis of common stock delistings. Journal of financial and quantitative analysis, 25(2), pp. 261-272.

Sarkissian, S. & Schill, M., 2003. The Overseas Listing Decision: New Evidence of Proximity Preference. The review of financial studies, 14(3), pp.

Sarkissian, S. & Schill, M., 2003. The Overseas Listing Decision: New Evidence of Proximity Preference. The review of financial studies, 14(3), pp.

In document Prediction of market switching and delisting events from OMX First North Nordic multilateral stock exchange (sivua 80-96)